Content uploaded by Nicolas Reiminger
Author content
All content in this area was uploaded by Nicolas Reiminger on Jun 16, 2020
Content may be subject to copyright.
DOI : 10.1016/j.jweia.2019.104032
1/29
CFD evaluation of mean pollutant concentration variations in step-down
1
street canyons as a function of aspect ratio
2
Nicolas Reiminger1,2*, José Vazquez2, Nadège Blond3, Matthieu Dufresne1, Jonathan Wertel1
3
1AIR&D, 67400, Illkirch-Graffenstaden, France
4
2ICUBE Laboratory, CNRS/University of Strasbourg, 67000, Strasbourg, France
5
3LIVE Laboratory, CNRS/University of Strasbourg, 67000, Strasbourg, France
6
*Corresponding author: Tel. +33 (0)3 69 06 49 40, Mail. nreiminger@air-d.fr
7
8
Please cite this paper as : Reiminger, N., Vazquez, J., Blond, N., Dufresne, M., Wertel, J., 2020. CFD
9
evaluation of mean pollutant concentration variations in step-down street canyons. Journal of Wind
10
Engineering and Industrial Aerodynamics 196, 104032. DOI: 10.1016/j.jweia.2019.104032
11
12
Abstract:
13
Atmospheric pollution became a big issue in densified urban areas where the ventilation in
14
streets is not sufficient. It is particularly the case for street surrounded by high buildings so-
15
called street canyons. The ventilation and, thus, the concentrations in this kind of street are
16
highly relying on geometric properties of the street (width of the street, heights of the buildings,
17
etc.) A Reynolds-averaged Navier-Stokes model is used to investigate the impact of two
18
geometric street ratios on pollutant dispersion: the ratio of the leeward to the windward building
19
height (H1/H2) and the ratio of the street width to the windward building height (W/H2). The
20
aim is to quantitatively assess the evolution of mean pollutant concentrations in the case of step-
21
down street canyons with H1/H2 ranging from 1.0 to 2.0 and street width ratios W/H2 ranging
22
from 0.6 to 1.4. Three types of recirculation regimes could be established, depending on the
23
number and the direction of the vortices occurring inside and outside the canyon. Evolution of
24
pollutant concentrations as a function of both ratios is provided as well as the recommended
25
regimes in the perspective of reducing pollutant concentration in step-down street canyons at
26
pedestrian level and near building faces.
27
Keywords: Air quality, Computational fluid dynamics, Street Canyon, Aspect ratio, Building
28
characteristics
29
DOI : 10.1016/j.jweia.2019.104032
2/29
1. Introduction
30
Air quality has become a major concern, especially in urban areas where air pollutant sources
31
are numerous and population density is high. Air quality is influenced by traffic-related
32
emissions and the local atmospheric environment which is highly dependent on street geometry.
33
Indeed, narrow streets surrounded by high buildings are more often subject to high pollutant
34
concentrations than wide streets with lower building heights, due to poorer ventilation. An
35
estimation of pollutant concentrations in streets depending on building configurations could
36
help urban planners to understand the impacts of street geometry on air quality and provide
37
keys to making suitable choices to lessening air pollution levels, as one of the key point
38
discussed by Bibri and Krogstie (2017) in order to achieve smart sustainable cities of the future.
39
The effects of street geometry on pollutant dispersion have already been studied extensively
40
with both experimental (Gerdes and Olivari, 1999; Hotchkiss and Harlow, 1973; Pavageau and
41
Schatzmann, 1999; Vardoulakis et al., 2003) and numerical methods (Aristodemou et al., 2018;
42
Bijad et al., 2016; Santiago and Martin, 2005; Tominaga and Stathopoulos, 2017; Vardoulakis
43
et al., 2003) and also at full-scale with in situ measurements (Qin and Kot, 1993; Vardoulakis
44
et al., 2002). Some authors have even studied the effects of roof shape on pollutant dispersion
45
(Takano and Moonen, 2013; Wen and Malki-Epshtein, 2018). However, most of these works
46
were conducted in symmetrical street canyons using buildings with the same height. Indeed,
47
streets surrounded by buildings of the same height do exist although streets with different
48
building heights, so-called asymmetrical street canyons, are found more often. Addepalli and
49
Pardyjak (2015) studied cases of step-down street canyons with a taller building on the leeward
50
side and showed that there are significant modifications of flow patterns depending on building
51
height and street width ratios. Xiaomin et al. (2006) performed a similar work with different
52
kinds of streets, including deep and wide symmetrical streets and step-up and step-down
53
asymmetrical streets, and showed that there are three major types of regimes in street canyons
54
DOI : 10.1016/j.jweia.2019.104032
3/29
depending on height and width ratios, especially in the case of step-down street canyons. In
55
spite of the several studies already done, and although there is a need for urban planners and
56
decision makers, quantitative information on how concentrations evolve with the modification
57
of street geometry is still lacking. Thus, further work is required in this direction.
58
The aim of this work is to provide information on how mean pollutant concentrations evolve
59
quantitatively in the case of step-down street canyons according to two specific ratios: the ratio
60
of the leeward building height to the windward building height H1/H2 and the ratio of the street
61
width to the windward building height W/H2, determined by computational fluid dynamics
62
(CFD) simulations. Section 2 presents the numerical model used in this work with the governing
63
equations, the boundary conditions and the numerical settings. Section 3 presents the validation
64
of the model versus experimental data in which a mesh sensitivity test and an evaluation of the
65
best turbulent Schmidt number are carried out. Finally, section 4 describes the results of the
66
study for several mean concentrations and a discussion of the results is proposed in section 5.
67
68
2. Numerical model
69
2.1. Computational domain and boundary conditions
70
Fig. 1 shows the computational domain of the street canyon, the dimensions of interest, the
71
localization of the different boundary conditions and the domain size.
72
In this study, H1 corresponds to the height of the leeward building, H2 corresponds to the height
73
of the windward building, W corresponds to the width between the two buildings and L
74
corresponds to the length of the street. Here, we study the case of a long canyon (L/W>5) with
75
the assumption that the interactions in the y-direction are negligible. To ensure this assumption
76
a 3D simulation was computed for this study, and the results were compared to 2D results.
77
Using a street canyon with L/W=10, it was found that the differences between 2D and 3D
78
simulation are fewer than 8% for |y|≤3H with y=0H the center plane of the street. For
79
DOI : 10.1016/j.jweia.2019.104032
4/29
3H<L/W<5H, differences are still acceptable but can reach 20%. According to this results, all
80
simulations were done in 2D in order to reduce calculation costs.
81
We followed the recommendations given by Franke et al. (2007) concerning the boundary
82
conditions and the domain size: the inlet boundary is placed 7×H2 away from the canyon; a
83
symmetry condition is applied at the top and the lateral boundaries, with the top placed 6×H2
84
away from the roofs of the buildings; the outlet boundary is placed 15×H2 away from the street
85
to allow for flow development using a freestream outlet, and no-slip conditions were applied to
86
all the other boundaries (roofs/walls of the buildings and the ground).
87
88
Fig. 1. Sketch of the computational domain
89
2.2. Governing equations
90
CFD simulations were carried out in OpenFOAM 5.0. Since in real contexts, full steady state is
91
not always reached, all the simulations were performed using the unsteady pimpleFoam solver
92
which is able to capture time instabilities. Reynolds-averaged Navier-Stokes (RANS)
93
methodology was used to solve the continuity and the momentum equations throughout the
94
computational domain by considering air as an incompressible fluid. This assumption can be
95
made because of the low wind velocities (<10m/s) which give low Mach numbers. The
96
corresponding continuity (1) and momentum (2) equations are given below:
97
(1)
98
DOI : 10.1016/j.jweia.2019.104032
5/29
(2)
99
where and
are the ith mean and the fluctuating velocities, respectively, is the ith
100
Cartesian coordinate,
is the mean pressure and is the kinematic viscosity.
101
Using RANS to solve turbulent flows requires choosing a turbulence model to solve the
102
Reynolds stress tensor
(3). The RNG k-ε model proposed by Yakhot et al. (1992) was
103
chosen for turbulent closure because the numerical results fitted well with the experimental data
104
(see section 3.1.). The corresponding equations for turbulent kinetic energy (4) and turbulent
105
dissipation rate (5) of the RNG model are given below. Taking R=0 and using the correct
106
constants, these equations also correspond to the standard k-ε model.
107
(3)
108
(4)
109
(5)
110
(6)
111
(7)
112
where and the mean strain tensor, is the ith mean velocity, is the
113
ith Cartesian coordinate, is the kinematic viscosity, k is the turbulent kinetic energy, is the
114
turbulent dissipation rate, is the Kronecker delta and is the turbulent viscosity. All the
115
other parameters are model constants given in Table 1 for both the standard and the RNG k-ε
116
models.
117
DOI : 10.1016/j.jweia.2019.104032
6/29
Table 1. Turbulence model constant values
118
Model
Cµ
Cε1
Cε2
σk
σε
η0
β
Standard k-ε
0.09
1.45
1.9
1.0
1.3
-
-
RNG k-ε
0.085
1.42
1.68
0.72
0.72
4.38
0.015
119
Pollutants are considered as passive scalars since no chemical effects are solved in this study.
120
The equation governing advection-diffusion for the passive pollutant dispersion given in
121
OpenFOAM was modified to take into account turbulent diffusivity. The corresponding
122
equation is given below:
123
(7)
124
where C is the pollutant concentration, is the molecular diffusion coefficient, is the
125
turbulent Schmidt number and is the source term of the pollutants (emissions).
126
The ratio
corresponds to the turbulent diffusion coefficient. The value of is constant
127
throughout the computational domain and fixed at 0.2. This value was chosen for the validation
128
step (see section 3.2.).
129
2.3. Numerical settings
130
Second order schemes were adopted for all the gradient, divergent and Laplacian terms. In
131
particular, for the Laplacian terms we used the ‘Gauss linear corrected’-scheme which is an
132
unbounded second order conservative scheme, the second order ‘Gauss linear’-scheme for the
133
gradient terms and the ‘Gauss linearUpwind’-scheme for the divergent terms, the latter scheme
134
being an unbounded upwind second order scheme.
135
All the simulations were run until the convergence was reached. To ensure the convergence of
136
the simulations, the values of the streamwise velocity U and the pollutant concentration C were
137
monitored for several points all over the canyon. Since all the simulations reached steady-state,
138
DOI : 10.1016/j.jweia.2019.104032
7/29
they were stopped when the values monitored were constant over time. Moreover, at the end of
139
the simulations all the residuals were under 10-5.
140
3. Model validation
141
The model was validated versus the experimental wind tunnel data proposed by Soulhac et al.
142
(2001). This experiment setup consists of a regular street canyon with H1/H2=1 and W/H2=1
143
with a gas released continuously at the center of the street. A summary of the boundary
144
conditions used for this validation is given in Table 2. A comparison between experimental and
145
numerical streamwise velocity was made to evaluate mesh sensitivity; another comparison
146
between experimental and numerical pollutant concentrations was made to find the turbulent
147
Schmidt number which gave the best results compared to the experiment.
148
149
Table 2. Summary of the boundary conditions
150
Inlet
Experimental velocity profile which corresponds to a power law profile with
, where =5.54m/s is the velocity at , =0.63m is the reference height,
=0.127 is the power law exponent and z the height from the ground.
, with
the turbulent intensity, with the Reynolds
number where U=4.43m/s is the mean inlet velocity, H=0.6m is the injection height and =1.56.10-
5 is the kinematic viscosity.
with =0.085 the CFD constant, and the turbulence length taken as equal to the
injection height (0.6m).
Outlet
Freestream outlet
Top
Symmetry plane
Lateral surfaces
Symmetry plane
Ground and
building surfaces
No slip condition (U=0m/s)
Emission
Line source with emission rate qm=1.10-4 µg/s localized at the middle of the street
151
152
DOI : 10.1016/j.jweia.2019.104032
8/29
3.1. Mesh sensitivity
153
Mesh sensitivity tests were carried out and compared to the experimental streamwise velocity
154
results to find the best compromise between the precision of the numerical results and
155
calculation costs.
156
Fig. 2. shows this comparison for three localized velocity profiles: on the leeward side of the
157
street (x/H=-0.2), in the middle of the street (x/H=0.0) and on the windward side of the street
158
(x/H=0.2). Three mesh-dependent results are proposed and the grid expansion ratio between
159
the coarse and the medium grid and between the medium and the fine grid is 2. Velocities and
160
heights are proposed in dimensionless form, corresponding to U/Umax with Umax=5m/s and z/H
161
with H=0.1m, respectively.
162
The results show good agreement between the experimental and numerical data whatever the
163
mesh refinement considered. There is a noticeable difference in the numerical results between
164
the coarse and the medium mesh in the street canyon (z/H<1). The difference between the
165
medium and the fine meshes is almost imperceptible apart from the low heights for which the
166
fine mesh results are closer to the experimental results. Thus, in the light of these results, the
167
fine mesh grid was adopted.
168
DOI : 10.1016/j.jweia.2019.104032
9/29
169
Fig. 2. Vertical distribution of numerical streamwise velocities for different mesh refinements compared to Soulhac et al.
170
(2001) experimental data
171
172
An additional mesh sensitivity study was performed on the variable of interest C, the pollutant
173
concentration, using the Grid Convergence Index (GCI) methodology proposed by Roach
174
(1994). This methodology is used to assess the mesh-related errors of a given mesh grid in view
175
of the fine and coarse grid results and depending on the grid expansion ratio and the order of
176
the numerical scheme used. The GCI for fine mesh grid error evaluation is given below:
177
(8)
178
where and are the results using the fine and coarse grid, respectively (here and
179
), r is the grid expansion ratio between the fine and the coarse grid and p is the
180
order of the numerical scheme.
181
The grid convergence index for the fine grid was calculated for 370 points uniformly distributed
182
in the street canyon with p = 2 (second order schemes) and r = 4 (the fine mesh is four times
183
DOI : 10.1016/j.jweia.2019.104032
10/29
smaller than the coarse mesh). The corresponding mean is 2% and the maximum
184
4%, thus corresponding to a sufficient grid resolution. The typical dimension of the chosen cells
185
is 0.0125 × H2.
186
3.2. Turbulent Schmidt number
187
According to Tominaga and Stathopoulos (2007), the optimal values of the turbulent Schmidt
188
number are widely spread between 0.2 and 1.3 and have a considerable influence on
189
pollutant mass transfer. Thus, must be chosen with care. To make this choice, several
190
simulations were performed for 0.1<<0.7 with steps of 0.1 and the results were compared
191
with the experimental data.
192
Fig. 3. shows the results for three localized concentration profiles: close to the leeward building
193
(x/H=-0.4), in the middle of the street (x/H=0.0) and close to the windward building (x/H=0.4).
194
The three closest numerical results compared to the experiment are shown and differ only by
195
the turbulent Schmidt number used: 0.1, 0.2 and 0.3. Concentrations and heights are proposed
196
in dimensionless form. The same dimensionless form as before was used for the heights (z/H)
197
and the dimensionless concentration was obtained using (9).
198
(9)
199
where C* is the dimensionless concentration, C is the concentration, UH is the velocity just over
200
the windward building, H2 is the windward building height, L is the pollutant injection length
201
and qm is the pollutant emission rate.
202
DOI : 10.1016/j.jweia.2019.104032
11/29
203
Fig. 3. Vertical distribution of numerical dimensionless concentrations for different Sct compared to Soulhac et al. (2001)
204
experimental data
205
206
The results show good agreement between the numerical and experimental data for =0.2.
207
Regarding this turbulent Schmidt number, for the leeward side there is generally an
208
overestimation of the concentrations in the upper part of the street and an underestimation in
209
the lower part of street while there is a general underestimation for the windward side. The
210
numerical results are less accurate with =0.1 and =0.3, so the value of 0.2 was kept for
211
the rest of the study. Using this turbulent Schmidt number, the mean normalized absolute error
212
over the experimental profiles was 10%. The corresponding 95th percentile was less than 30%
213
and the maximal differences between the experimental and numerical results occurred near the
214
ground.
215
The models used in the present paper (RANS and RNG k-ε) give a global underestimation of
216
the turbulent momentum diffusion leading to low turbulent Sct. The turbulent Schmidt number
217
taken as 0.2 is in coherence with other authors results who took a low Sct as 0.3 for the same
218
models (Tominaga and Stathopoulos, 2007). It should be noted that the value of 0.2 could not
219
be the best for all the geometric ratios considered in this work. However, it was decided to
220
always use the same Sct in the whole study, which is a common practice done by the scientific
221
DOI : 10.1016/j.jweia.2019.104032
12/29
community (Takano and Moonen, 2013 ; Wen and Malki-Epshtein, 2018 ; Cui et al., 2016), in
222
order to only compare the influence of the geometric properties of the buildings on the mean
223
concentrations and to avoid multi parameter comparisons.
224
4. Effects of street dimensions on mean concentrations
225
Exactly the same conditions as defined previously were used for the present study, except for
226
the geometric properties of the street and in particular H1 and W. To study the mean
227
concentrations in the street canyon, several couples of height ratios H1/H2 and width ratios
228
W/H2 were considered. The present work is limited to a step-down street canyon configuration
229
where H1/H2>1.0. The following height ratios were used: 1.0, 1.2, 1.4, 1.6, 1.8 and 2.0. For
230
each of these height ratios, 5 width ratios were considered: 0.6, 0.8, 1.0, 1.2 and 1.4, giving a
231
total number of 30 simulations and an overall idea of how could evolve mean concentrations in
232
step-down street canyons. This number does not include certain particular cases that were also
233
simulated when the results were strongly different between two cases (e.g. when for a given
234
width ratio, two successive height ratios results in two different regimes). A case table of all the
235
ratios considered in this work is proposed in Table 3.
236
Table 3. Case table of all geometric ratios considered (● : couples of ratios initially considered, ○ : specific cases considered
237
aftermath)
238
W/H2
0.6
0.8
1.0
1.2
1.4
H1/H2
2.0
●
●
●
●
●
1.9
1.8
●
●
●
●
●
1.7
1.6
●
●
●
●
●
1.5
○
1.4
●
●
●
●
●
1.3
○
○
○
1.2
●
●
●
●
●
1.1
○
○
1.0
●
●
●
●
●
239
DOI : 10.1016/j.jweia.2019.104032
13/29
Fig. 4 shows the localization of the mean concentrations studied in this paper. Here, we study:
240
- The concentration averaged all over the street (in the W×H2 area),
241
- The mean concentration on a vertical profile placed 0.1H2 from the windward building
242
facade (concentration averaged for the H2 height) and another vertical profile placed 0.1H2
243
from the leeward building facade (concentration averaged for the H2 height). These mean
244
concentrations are relevant for people living in the buildings near the street.
245
- The mean concentration for a horizontal profile placed 0.1H2 from the ground
246
(concentration averaged for the W length). This mean concentration is relevant for
247
pedestrians in the street.
248
249
250
Fig. 4. Localization of the mean concentrations studied.
251
252
All the concentrations will be given in dimensionless form. The dimensioned concentrations
253
could also be retrieved using (9) with =2.75m/s, H2=0.1m, L=0.0025m and =1.10-4 µg/s.
254
4.1.Vorticity and recirculation regimes in the street canyon
255
Flow velocities and recirculation patterns have a significant impact on pollutant dispersion and
256
thus on pollutant concentrations inside and outside the street canyon. The modifications of flow
257
DOI : 10.1016/j.jweia.2019.104032
14/29
velocities and recirculation patterns are caused solely by the geometric properties of the street
258
(H1/H2 and W/H2) as all the simulations were run using the same velocity inlet profile.
259
Out of the total number of simulations performed, three types of recirculation regimes were
260
found. Fig. 5. shows an example of each regime with the velocity vectors and the corresponding
261
y-vorticity given by equation (10). These three regimes stand out due to their number of
262
recirculation zones inside and outside the canyon.
263
(10)
264
Regime A corresponds to a big single vortex localized in the canyon. For this regime, vorticity
265
is globally positive in the canyon, which means that the vortex rotates clockwise. Regime B
266
corresponds to two vortices, one large vortex in the canyon and a second localized mostly over
267
the canyon and the windward building. The large vortex in the canyon is very similar to that of
268
regime A, but here the vorticity is mostly negative, and the vortex rotates counterclockwise.
269
The second vortex localized outside the canyon rotates clockwise. Regime C corresponds to
270
three vortices, two contra-rotative vortices localized in the canyon and the third vortex mostly
271
localized over the windward building. This regime appears to be a combination of regimes A
272
and B, with the clockwise-vortex of regime A in the low part of the street and the
273
counterclockwise-vortex of regime B situated just over it. The same clockwise-outside-vortex
274
of regime B is also observed.
275
Xiaomin et al. (2006) gave the critical value of H1/H2 for several W/H2 corresponding to the
276
limit between regime A and regime B/C without distinction between B and C. Their results are
277
compared with those of the present study for W/H2 from 0.6 to 1.4 and are shown in Fig. 6. with
278
the gray area corresponding to the switching area between regime A and regime B/C. The
279
boundary conditions were the same between both studies.
280
281
DOI : 10.1016/j.jweia.2019.104032
15/29
282
Fig. 5. Recirculation patterns, velocity vectors and y-vorticity for different geometric ratios H1/H2 and W/H2
283
284
The results obtained after the simulations showed a trend similar to that of the results of
285
Xiaomin et al. (2006). The critical value of H1/H2 increases when the distance between the
286
buildings increases and the zone of change between regime A and regime B/C is quite similar
287
for both studies. However, critical values seem to be reached sooner according to our results
288
(i.e. for smaller H1/H2) with a maximal difference of 0.1 compared to the results of Xiaomin et
289
al.
290
Some simulations were rerun using the turbulent conditions of Xiaomin’s et al., that is, using
291
the standard k-ε turbulent closure. The results, also presented in Fig. 6., show this time perfect
292
concordance between both studies. Thus, turbulent closure schemes have an influence on the
293
critical values of H1/H2. This difference between critical values when using standard k-ε or
294
RNG k-ε are, however, quite small with a maximum difference of 0.1 for the ratio H1/H2.
295
DOI : 10.1016/j.jweia.2019.104032
16/29
296
Fig. 6. Comparison of regime changing zones between the present study and the results of Xiaomin et al. (2006) using RNG
297
and standard k-ε turbulent closure.
298
299
4.2. Impact of the regimes on pollutant dispersion
300
Three examples of pollutant dispersion in the street canyon for each regime are shown in Fig. 7.
301
The overall concentrations in the street canyons being very different between the three regimes,
302
the color scale is different for each of them. The velocity vectors are provided in order to better
303
understand the differences in the concentration fields for the three regimes.
304
The evolution of the concentration field, the overall magnitude of concentration, and the most
305
impacted building are directly linked with the type of regime being established. In regime A,
306
the pollutants released at ground level are mostly dispersed towards the leeward building due
307
to the single clockwise vortex established in the street. In regime B, the apparition of a second
308
vortex due to the increase of the leeward building height and the decrease of the distance
309
between building leads to a change in the dispersion of pollutants. The vortex in the street being
310
in this case counter clockwise, the most impacted building became the windward building.
311
Moreover, concentrations are overall higher in this case and it seems to be the consequence of
312
DOI : 10.1016/j.jweia.2019.104032
17/29
the clockwise vortex localized just above which is driving a part of the pollutants which left the
313
street to the street again. For the last regime, regime C, both buildings are highly impacted. The
314
difference with the regime B is not only the apparition of a third vortex, but the fact that two
315
vortices are localized in the street between the buildings. Due to this two vortices, the pollutants
316
released at ground level are dispersed to the leeward building but, because of the second vortex
317
in the canyon, they are more homogenized in the low part of the street and seem to be more
318
stagnant. It should also be noted that global velocities in the street tend to decrease with the
319
increase of the leeward building height and the decrease of the distance between building which
320
also conduct to higher pollutant concentrations.
321
322
Fig. 7. Three examples of dimensionless concentrations in a street canyon for each type of regime.
323
DOI : 10.1016/j.jweia.2019.104032
18/29
4.3. Mean concentration in the street canyon
324
Initially, the results were studied by considering the mean concentrations of the whole street.
325
Fig. 8 shows the dimensionless street averaged concentrations (i.e. the mean concentration of
326
the W×H2 surface) proposed for several H1/H2 and W/H2 ratios and the different types of
327
regime are also specified.
328
329
330
Fig. 8. Dimensionless street averaged concentrations according to the ratio H1/H2 and W/H2
331
332
The results show that the evolution of mean concentrations is highly dependent on the type of
333
regime in place. The mean concentrations are indeed highest when regime C is in place and
334
lowest when regime A is in place.
335
In regime A, for a given distance between buildings (i.e. a given W/H2), the mean
336
concentrations are the same whatever the height of the leeward building. Thus, only the distance
337
between buildings has an impact on the mean concentrations in the street. For a fixed leeward
338
building height, the mean concentrations in the street increase when the distance between
339
buildings decrease. This increase is not constant and becomes higher when ratio W/H2
340
DOI : 10.1016/j.jweia.2019.104032
19/29
decreases. For example, the mean concentration increases by 23% between W/H2=1.2 and
341
W/H2=1.0 and then by 37% between W/H2=1.0 and W/H2=0.8. Lastly, for the H1/H2 and
342
W/H2 ratios studied in this work, the factor between the lowest and the highest mean
343
concentration for regime A is equal to 2.
344
In regime B, the evolution of the mean street concentrations is dependent on both ratios H1/H2
345
and W/H2: for a given leeward building height, the mean street concentrations increase when
346
the distance between the buildings decreases; for a given distance between buildings, the mean
347
concentration increases when the leeward building height increases. In addition, the increases
348
between mean concentrations are not constant and become higher when H1/H2 increases and
349
W/H2 decreases. The factor between the highest and lowest mean concentrations in the case of
350
regime B is around 5.
351
In regime C, the evolution of the street mean concentrations is also dependent on both ratios
352
H1/H2 and W/H2 but is no longer monotonous. Indeed, for a given distance between the
353
buildings, the mean street concentrations first increase and then become constant. If the leeward
354
building height is high enough, this mean concentration can then decrease. In this third case, a
355
maximal mean concentration is reached. Mean street concentrations are highest for this regime
356
with, in the worst-case concentrations, 50 times that of the regular case H1/H2=W/H2=1.0.
357
Lastly, considering the whole series of simulations run in this study, for a given H1/H2 ratio,
358
the mean concentrations increase as the distance between buildings decreases, whatever the
359
three regimes observed. The evolution of the mean concentrations for a given W/H2 is
360
nevertheless dependent on the regime.
361
362
363
364
DOI : 10.1016/j.jweia.2019.104032
20/29
4.4. Mean concentration on the building sides
365
The results were then studied considering only the windward and the leeward building sides.
366
Fig. 9 shows the dimensionless windward side averaged concentrations (i.e. the mean
367
concentrations averaged over the windward profile) proposed for several H1/H2, and W/H2
368
ratios and the different types of regime are also specified. Fig. 10 gives the same information,
369
but considering the dimensionless, averaged leeward side concentrations (i.e. the mean
370
concentrations averaged over the leeward profile).
371
As can be seen in Fig. 9 and Fig. 10, the evolution of the mean concentrations on the two
372
building sides are similar. However, the mean concentrations could be higher or lower on the
373
windward side, depending on the recirculation regimes.
374
In Regime A, for a given distance between buildings (i.e. a given W/H2 ratio), the mean leeward
375
and windward concentrations are constant whatever the H1/H2 ratio. However, the mean
376
concentration values are different, with concentrations globally twice as high on the leeward
377
side. This observation is linked to the characteristics of regime A described in section 4.1.
378
Indeed, for all the cases in which regime A occurs, a large clockwise rotating vortex appears
379
which spreads the pollutants released at ground level to the leeward side.
380
In regime B, the mean concentrations are no longer constant for a given distance between
381
buildings but depend on both ratios H1/H2 and W/H2. This time the mean concentrations are
382
higher on the windward side according to the counterclockwise vortex occurring in regime B,
383
which spreads the pollutants released at ground level to the windward side. The mean
384
concentrations on the windward side are globally three times higher than those of the leeward
385
side.
386
In regime C, the mean concentrations still depend on both ratios H1/H2 and W/H2 and the
387
concentrations are much higher than in regime B. The mean concentrations are globally higher
388
DOI : 10.1016/j.jweia.2019.104032
21/29
on the leeward side but this is not always true. Indeed, for H1/H2=2.0 and W/H2=0.8, the mean
389
windward concentration is higher. It is much more difficult to interpret this difference than
390
those of the two previous regimes because two vortices are localized in the canyon in this case.
391
However, in this case the vortex is clockwise and localized near the emission source. The
392
pollutants released near the ground are thus initially spread to the leeward side and it is only
393
afterwards that the second vortex spreads them to the windward side. This explains why the
394
mean concentrations are mostly higher on the leeward side than on the windward side.
395
Finally, if we focus on how the mean concentrations evolve when the regimes change (e.g.
396
when switching from regime A to regime B), there is a notable difference between the windward
397
and leeward sides. Indeed, for a switch from regime A to regime B, whereas the mean
398
concentrations increase by a factor 6 on the windward side, the concentrations on the leeward
399
side are almost equal. Moreover, on the leeward side, the mean concentration observed in the
400
case of regime B did not increase much when H1/H2 increased or W/H2 decreased compared
401
to the windward side.
402
403
Fig. 9. Dimensionless windward profile averaged concentrations according to the ratios H1/H2 and W/H2.
404
DOI : 10.1016/j.jweia.2019.104032
22/29
405
Fig. 10. Dimensionless leeward profile averaged concentrations according to the ratios H1/H2 and W/H2.
406
407
4.5.Mean concentration at ground level
408
Finally, the results were studied at ground level and Fig. 11 shows the dimensionless ground
409
averaged concentrations (i.e. the mean concentrations averaged over the ground profile)
410
proposed for several H1/H2 and W/H2 ratios; the different types of regime are also specified.
411
At ground level, the evolution of mean concentrations is similar for the leeward profile and the
412
whole street: regime A leads to constant mean concentrations for a given distance between
413
buildings; regime B leads to mean concentrations depending on both the distance between
414
buildings and difference in height between the two buildings; regime C leads to the same
415
observation as regime B, the difference being that for a given distance between buildings, a
416
maximal mean concentration is reached, after which this concentration decreases with the
417
increase in the difference in height between the two buildings.
418
DOI : 10.1016/j.jweia.2019.104032
23/29
419
Fig. 11. Dimensionless ground profile averaged concentrations according to ratio H1/H2 and W/H2.
420
421
5. Discussion
422
Choices were made regarding the turbulence model used as well as the isothermal assumption
423
taken to fulfil this work. These choices could affect the presented results and are worth
424
discussing about.
425
Based on comparison with experimental data, the RNG turbulence model was selected. This
426
model is an isotropic linear k-ε based model that is known to have some limitations for highly
427
transient cases, especially in a wake of a body, including flows behind the leeward walls of
428
street canyons. To avoid such problems, non-linear turbulence models or anisotropic models
429
such as the Reynolds Stress Model (RSM) should be used. However, these models are time
430
consuming and are more difficult to converge. In addition, they seem to give not as much
431
improvements as expected in the case of isolated buildings or street canyons. Indeed,
432
Papageorgakis and Assanis (1999) showed that the linear RNG k-ε model gives significant
433
improvements compared to the standard model for recirculatory flow such for backward facing
434
step cases. Moreover, according to the same authors, the non-linear RNG model is not very
435
DOI : 10.1016/j.jweia.2019.104032
24/29
attractive, yielding not to great improvements. Finally, Koutsouarakis et al. (2012) showed for
436
six street canyons with different aspect ratios that the RNG model gives the best performances
437
for each case compared to the standard model as well as compared to the RSM model.
438
The whole study was conducted considering neutral (isothermal) conditions since ambient and
439
wall temperatures were considered equal. Thus, only the forced convection due to the wind was
440
considered. More complex cases could appear when the building walls are heated by solar
441
radiations conducting to unstable conditions where natural convection appears. For this cases,
442
results in terms of recirculation regimes or pollutant concentrations can be different. Wang et
443
al. (2011) studied the cases of leeward, ground, and windward heated walls in a regular street
444
canyon and compared the results with the neutral case (without wall heating). They found that,
445
except for the case of the windward heated wall, the recirculation pattern in the street is always
446
the same. Concentrations are different depending on the case, but they are always lower than
447
for the neutral case. These results are confirmed by Allegrini et al. (2013) who did the same
448
work with several wind speed and also simulated a case where all walls are heated. This case
449
also leads to the same recirculation pattern as for the neutral case. According to these results, it
450
could be said that the results given in this study are not only good for one considering neutral
451
cases but are also a good first approximation of unstable cases. Pollutant concentrations being
452
greater for the neutral case than for the unstable case leading thus to a safer approach.
453
6. Conclusion
454
The effects of step-down street canyon geometric properties on recirculation patterns and mean
455
pollutant concentrations in a street were studied with a CFD model. This study considered 6
456
height ratios H1/H2 (from 1.0 to 2.0 with a 0.2 step) and 5 width ratios W/H2 (from 0.6 to 1.4
457
with a 0.2 step). The main conclusions are as follows:
458
(a) Three types of regimes can occur as a function of both the height and width ratios of the
459
street. Flow velocities and direction in the street, and thus pollutant concentrations,
460
DOI : 10.1016/j.jweia.2019.104032
25/29
depend heavily on the type of regime being established. The three types of regime were
461
characterized by the number of vortices established and their direction: regime A
462
corresponded to a single clockwise vortex in the canyon; regime B corresponded to a
463
counter-clockwise vortex in the canyon and a clockwise vortex over the windward
464
building; regime C corresponded to two contra-rotating vortices in the canyon and a
465
clockwise vortex over the windward building.
466
(b) The critical values of H1/H2 corresponding to a change in the type of regime for a given
467
width ratio were determined. The critical values obtained were differed as a function of
468
the turbulence closure scheme used. These differences were never greater than 0.1 when
469
using standard or RNG k-epsilon turbulence schemes.
470
(c) Whatever the mean concentration considered (in the whole canyon, at pedestrian level
471
or near the building faces), the mean concentrations were lowest in the case of regime
472
A and highest in the case of regime C. Regime B therefore corresponded to an
473
intermediary state.
474
(d) The mean concentrations increased globally as differences in building height increased
475
(H1/H2 ratio), and with the decrease of street width (W/H2), except for the case of
476
regime A where the evolutions of mean concentrations depended only on street width.
477
(e) The quantitative evolution of the mean pollutant concentration in the whole street at
478
pedestrian level and near the building faces was proposed.
479
As a summary, in order to have a good ventilation in step-down street canyons and in the
480
perspective of reducing mean pollutant concentration of the whole street at pedestrian level and
481
near building faces, we recommend choosing carefully the height ratio H1/H2 as well as the
482
width ratio W/H2 in order to be in the case of a regime A.
483
These conclusions and results were obtained for a given type of street canyon and they should
484
be extended to consider other types such as step-up street canyons and wider and deeper
485
DOI : 10.1016/j.jweia.2019.104032
26/29
canyons. Moreover, these results were obtained considering flat roofs. However, this type of
486
roof is not the only kind of roof used for buildings and further works should be carried out to
487
obtain information on other types of roof.
488
489
Acknowledgments
490
We would like to thank the ANRT (Association Nationale de la Recherche et de la Technologie)
491
for their support.
492
493
References
494
Addepalli, B., Pardyjak, E.R., 2015. A study of flow fields in step-down street canyons.
495
Environmental Fluid Mechanics 15, 439–481. https://doi.org/10.1007/s10652-014-
496
9366-z
497
Allegrini, J., Dorer, V., Carmeliet, J., 2013. Wind tunnel measurements of buoyant flows in
498
street canyons. Building and Environment 59, 315–326.
499
https://doi.org/10.1016/j.buildenv.2012.08.029
500
Aristodemou, E., Boganegra, L.M., Mottet, L., Pavlidis, D., Constantinou, A., Pain, C., Robins,
501
A., ApSimon, H., 2018. How tall buildings affect turbulent air flows and dispersion of
502
pollution within a neighbourhood. Environmental Pollution 233, 782–796.
503
https://doi.org/10.1016/j.envpol.2017.10.041
504
Bibri, S.E., Krogstie, J., 2017. Smart sustainable cities of the future: An extensive
505
interdisciplinary literature review. Sustainable Cities and Society 31, 183–212.
506
https://doi.org/10.1016/j.scs.2017.02.016
507
DOI : 10.1016/j.jweia.2019.104032
27/29
Bijad, E., Delavar, M.A., Sedighi, K., 2016. CFD simulation of effects of dimension changes
508
of buildings on pollution dispersion in the built environment. Alexandria Engineering
509
Journal 55, 3135–3144. https://doi.org/10.1016/j.aej.2016.08.024
510
Cui, P.-Y., Li, Z., Tao, W.-Q., 2016. Buoyancy flows and pollutant dispersion through different
511
scale urban areas: CFD simulations and wind-tunnel measurements. Building and
512
Environment 104, 76–91. https://doi.org/10.1016/j.buildenv.2016.04.028
513
Franke, J., Hellsten, A., Schlünzen, H., Carissimo, B., 2007. Best practice guideline for the
514
CFD simulation of flows in the urban environment. COST Action 732.
515
Gerdes, F., Olivari, D., 1999. Analysis of pollutant dispersion in an urban street canyon. Journal
516
of Wind Engineering and Industrial Aerodynamics 82, 105–124.
517
https://doi.org/10.1016/S0167-6105(98)00216-5
518
Hotchkiss, R.S., Harlow, F.H., 1973. Air Pollution Transport in Street Canyons. National
519
Technical Information Service.
520
Koutsourakis, N., Bartzis, J.G., Markatos, N.C., 2012. Evaluation of Reynolds stress, k-ε and
521
RNG k-ε turbulence models in street canyon flows using various experimental datasets.
522
Environmental Fluid Mechanics 12, 379–403. https://doi.org/10.1007/s10652-012-
523
9240-9
524
Pavageau, M., Schatzmann, M., 1999. Wind tunnel measurements of concentration fluctuations
525
in an urban street canyon. Atmospheric Environment 33, 3961–3971.
526
https://doi.org/10.1016/S1352-2310(99)00138-7
527
Qin, Y., Kot, S.C., 1993. Dispersion of vehicular emission in street canyons, Guangzhou City,
528
South China (P.R.C.). Atmospheric Environment. Part B. Urban Atmosphere 27, 283–
529
291. https://doi.org/10.1016/0957-1272(93)90023-Y
530
Roache, P.J., 1994. Perspective: A Method for Uniform Reporting of Grid Refinement Studies.
531
Journal of Fluids Engineering 116, 405. https://doi.org/10.1115/1.2910291
532
DOI : 10.1016/j.jweia.2019.104032
28/29
Santiago, J.L., Martin, F., 2005. Modelling the air flow in symmetric and asymmetric street
533
canyons. International Journal of Environment and Pollution 25, 145.
534
https://doi.org/10.1504/IJEP.2005.007662
535
Soulhac, L., Mejean, P., Perkins, R.J., 2001. Modelling the transport and dispersion of
536
pollutants in street canyons. International Journal of Environment and Pollution 16, 404.
537
https://doi.org/10.1504/IJEP.2001.000636
538
Takano, Y., Moonen, P., 2013. On the influence of roof shape on flow and dispersion in an
539
urban street canyon. Journal of Wind Engineering and Industrial Aerodynamics 123,
540
107–120. https://doi.org/10.1016/j.jweia.2013.10.006
541
Tominaga, Y., Stathopoulos, T., 2017. Steady and unsteady RANS simulations of pollutant
542
dispersion around isolated cubical buildings: Effect of large-scale fluctuations on the
543
concentration field. Journal of Wind Engineering and Industrial Aerodynamics 165, 23–
544
33. https://doi.org/10.1016/j.jweia.2017.02.001
545
Tominaga, Y., Stathopoulos, T., 2007. Turbulent Schmidt numbers for CFD analysis with
546
various types of flowfield. Atmospheric Environment 41, 8091–8099.
547
https://doi.org/10.1016/j.atmosenv.2007.06.054
548
Vardoulakis, S., Fisher, B.E.A., Pericleous, K., Gonzalez-Flesca, N., 2003. Modelling air
549
quality in street canyons: a review. Atmospheric Environment 37, 155–182.
550
https://doi.org/10.1016/S1352-2310(02)00857-9
551
Vardoulakis, S., Gonzalez-Flesca, N., Fisher, B.E.A., 2002. Assessment of traffic-related air
552
pollution in two street canyons in Paris: implications for exposure studies. Atmospheric
553
Environment 36, 1025–1039. https://doi.org/10.1016/S1352-2310(01)00288-6
554
Wang, P., Zhao, D., Wang, W., Mu, H., Cai, G., Liao, C., 2011. Thermal Effect on Pollutant
555
Dispersion in an Urban Street Canyon. International Journal of Environmental Research
556
5, 813–820. https://doi.org/10.22059/ijer.2011.388
557
DOI : 10.1016/j.jweia.2019.104032
29/29
Wen, H., Malki-Epshtein, L., 2018. A parametric study of the effect of roof height and
558
morphology on air pollution dispersion in street canyons. Journal of Wind Engineering
559
and Industrial Aerodynamics 175, 328–341.
560
https://doi.org/10.1016/j.jweia.2018.02.006
561
Xiaomin, X., Huang, Z., Wang, J., 2006. The impact of urban street layout on local atmospheric
562
environment. Building and Environment 41, 1352–1363.
563
https://doi.org/10.1016/j.buildenv.2005.05.028
564
Yakhot, V., Orszag, S.A., Thangam, S., Gatski, T.B., Speziale, C.G., 1992. Development of
565
turbulence models for shear flows by a double expansion technique. Physics of Fluids
566
A: Fluid Dynamics 4, 1510–1520. https://doi.org/10.1063/1.858424
567
568
